Re: [PATCH v16 3/5] drm/xe/uapi: Define drm_xe_vm_get_property

[Jianxun Zhang <jianxun.zhang@intel.com>]

On 4/2/25 18:19, Jianxun Zhang wrote:
> 
> 
> On 4/1/25 15:26, Jonathan Cavitt wrote:
>> Add initial declarations for the drm_xe_vm_get_property ioctl.
>>
>> v2:
>> - Expand kernel docs for drm_xe_vm_get_property (Jianxun)
>>
>> v3:
>> - Remove address type external definitions (Jianxun)
>> - Add fault type to xe_drm_fault struct (Jianxun)
>>
>> v4:
>> - Remove engine class and instance (Ivan)
>>
>> Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
>> Cc: Zhang Jianxun <jianxun.zhang@intel.com>
>> Cc: Ivan Briano <ivan.briano@intel.com>
>> ---
>>   include/uapi/drm/xe_drm.h | 75 +++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 75 insertions(+)
>>
>> diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h
>> index 616916985e3f..45f7560879cb 100644
>> --- a/include/uapi/drm/xe_drm.h
>> +++ b/include/uapi/drm/xe_drm.h
>> @@ -81,6 +81,7 @@ extern "C" {
>>    *  - &DRM_IOCTL_XE_EXEC
>>    *  - &DRM_IOCTL_XE_WAIT_USER_FENCE
>>    *  - &DRM_IOCTL_XE_OBSERVATION
>> + *  - &DRM_IOCTL_XE_VM_GET_PROPERTY
>>    */
>>   /*
>> @@ -102,6 +103,7 @@ extern "C" {
>>   #define DRM_XE_EXEC            0x09
>>   #define DRM_XE_WAIT_USER_FENCE        0x0a
>>   #define DRM_XE_OBSERVATION        0x0b
>> +#define DRM_XE_VM_GET_PROPERTY        0x0c
>>   /* Must be kept compact -- no holes */
>> @@ -117,6 +119,7 @@ extern "C" {
>>   #define DRM_IOCTL_XE_EXEC            DRM_IOW(DRM_COMMAND_BASE + 
>> DRM_XE_EXEC, struct drm_xe_exec)
>>   #define DRM_IOCTL_XE_WAIT_USER_FENCE        
>> DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_WAIT_USER_FENCE, struct 
>> drm_xe_wait_user_fence)
>>   #define DRM_IOCTL_XE_OBSERVATION        DRM_IOW(DRM_COMMAND_BASE + 
>> DRM_XE_OBSERVATION, struct drm_xe_observation_param)
>> +#define DRM_IOCTL_XE_VM_GET_PROPERTY    DRM_IOWR(DRM_COMMAND_BASE + 
>> DRM_XE_VM_GET_PROPERTY, struct drm_xe_vm_get_property)
>>   /**
>>    * DOC: Xe IOCTL Extensions
>> @@ -1189,6 +1192,78 @@ struct drm_xe_vm_bind {
>>       __u64 reserved[2];
>>   };
>> +/** struct xe_vm_fault - Describes faults for 
>> %DRM_XE_VM_GET_PROPERTY_FAULTS */
>> +struct xe_vm_fault {
>> +    /** @address: Address of the fault */
>> +    __u64 address;
>> +    /** @address_precision: Precision of faulted address */
>> +    __u32 address_precision;
>> +    /** @access_type: Type of address access that resulted in fault */
>> +    __u8 access_type;
>> +    /** @fault_type: Type of fault reported */
>> +    __u8 fault_type;
>> +    /** @fault_level: fault level of the fault */
>> +    __u8 fault_level;
>> +    /** @pad: MBZ */
>> +    __u8 pad;
>> +    /** @reserved: MBZ */
>> +    __u64 reserved[4];
> Why are engine class and instance gone? I think we got agreement on not 
> reporting class 0x4, but the rest should be kept. If KMD can have these 
> in dmesg, why can't we report these in VK report?
Never mind. I just caught up with the offline traffic. I have no opinion 
on this.

Thanks.

>> +};
>> +
>> +/**
>> + * struct drm_xe_vm_get_property - Input of 
>> &DRM_IOCTL_XE_VM_GET_PROPERTY
>> + *
>> + * The user provides a VM and a property to query among 
>> DRM_XE_VM_GET_PROPERTY_*,
>> + * and sets the values in the vm_id and property members, 
>> respectively.  This
>> + * determines both the VM to get the property of, as well as the 
>> property to
>> + * report.
>> + *
>> + * If size is set to 0, the driver fills it with the required size 
>> for the
>> + * requested property.  The user is expected here to allocate memory 
>> for the
>> + * property structure and to provide a pointer to the allocated 
>> memory using the
>> + * data member.  For some properties, this may be zero, in which 
>> case, the
>> + * value of the property will be saved to the value member and size 
>> will remain
>> + * zero on return.
>> + *
>> + * If size is not zero, then the IOCTL will attempt to copy the 
>> requested
>> + * property into the data member.
>> + *
>> + * The IOCTL will return -ENOENT if the VM could not be identified 
>> from the
>> + * provided VM ID, or -EINVAL if the IOCTL fails for any other 
>> reason, such as
>> + * providing an invalid size for the given property or if the 
>> property data
>> + * could not be copied to the memory allocated to the data member.
>> + *
>> + * The property member can be:
>> + *  - %DRM_XE_VM_GET_PROPERTY_FAULTS
>> + */
>> +struct drm_xe_vm_get_property {
>> +    /** @extensions: Pointer to the first extension struct, if any */
>> +    __u64 extensions;
>> +
>> +    /** @vm_id: The ID of the VM to query the properties of */
>> +    __u32 vm_id;
>> +
>> +#define DRM_XE_VM_GET_PROPERTY_FAULTS        0
>> +    /** @property: property to get */
>> +    __u32 property;
>> +
>> +    /** @size: Size to allocate for @data */
>> +    __u32 size;
>> +
>> +    /** @pad: MBZ */
>> +    __u32 pad;
>> +
>> +    union {
>> +        /** @data: Pointer to user-defined array of flexible size and 
>> type */
>> +        __u64 data;
>> +        /** @value: Return value for scalar queries */
>> +        __u64 value;
>> +    };
>> +
>> +    /** @reserved: MBZ */
>> +    __u64 reserved[3];
>> +};
>> +
>>   /**
>>    * struct drm_xe_exec_queue_create - Input of 
>> &DRM_IOCTL_XE_EXEC_QUEUE_CREATE
>>    *
>